《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (7): 2184-2191.DOI: 10.11772/j.issn.1001-9081.2021071319

• 多媒体计算与计算机仿真 • 上一篇    

基于残差注意力机制的点云配准算法

秦庭威1,2, 赵鹏程1,2, 秦品乐1,2(), 曾建朝1,2, 柴锐1,2, 黄永琦1,2   

  1. 1.山西省医学影像人工智能工程技术研究中心(中北大学),太原 030051
    2.中北大学 大数据学院,太原 030051
  • 收稿日期:2021-07-22 修回日期:2021-10-13 接受日期:2021-10-18 发布日期:2021-11-01 出版日期:2022-07-10
  • 通讯作者: 秦品乐
  • 作者简介:秦庭威(1997—),男,陕西渭南人,硕士研究生,CCF会员,主要研究方向:点云配准、机器学习
    赵鹏程(1995—),男,陕西渭南人,硕士,CCF会员,主要研究方向:三维点云处理、计算机视觉
    曾建朝(1963—),男,山西太原人,教授,博士,CCF会员,主要研究方向:复杂系统的维护决策和健康管理
    柴锐(1985—),男,山西运城人,讲师,博士,CCF会员,主要研究方向:医学影像处理
    黄永琦(1997—),男,山西太原人,硕士研究生,CCF会员,主要研究方向:点云配准、计算机视觉。
  • 基金资助:
    山西省重点研发计划项目(201803D31212-1);山西省工程技术研究中心建设项目(201805D121008)

Point cloud registration algorithm based on residual attention mechanism

Tingwei QIN1,2, Pengcheng ZHAO1,2, Pinle QIN1,2(), Jianchao ZENG1,2, Rui CHAI1,2, Yongqi HUANG1,2   

  1. 1.Shanxi Medical Imaging and Data Analysis Engineering Research Center (North University of China),Taiyuan Shanxi 030051,China
    2.College of Data Science and Technology,North University of China,Taiyuan Shanxi 030051,China
  • Received:2021-07-22 Revised:2021-10-13 Accepted:2021-10-18 Online:2021-11-01 Published:2022-07-10
  • Contact: Pinle QIN
  • About author:QIN Tingwei, born in 1997, M. S. candidate. His research interests include point cloud registration, machine learning.
    ZHAO Pengcheng, born in 1995, M. S. His research interests include 3D point cloud processing, computer vision.
    ZENG Jianchao, born in 1963, Ph. D., professor. His research interests include maintenance decision and health management of complex system.
    CHAI Rui, born in 1985, Ph. D., lecturer. His research interests include medical image processing.
    HUANG Yongqi, born in 1997, M. S. candidate. His research interests include point cloud registration, computer vision.
  • Supported by:
    Shanxi Provincial Key Research and Development Plan(201803D31212-1);Construction Project of Engineering Technology Research Center of Shanxi Province(201805D121008)

摘要:

针对传统点云配准算法精度低、鲁棒性差以及放疗前后癌症患者无法实现精确放疗的问题,提出一种基于残差注意力机制的点云配准算法(ADGCNNLK)。首先,在动态图深度卷积网络(DGCNN)中添加残差注意力机制来有效地利用点云的空间信息,并减少信息损失;然后,利用添加残差注意力机制的DGCNN提取点云特征,这样做不仅可以在保持点云置换不变性的同时捕捉点云的局部几何特征,也可以在语义上将信息聚合起来,从而提高配准效率;最后,将提取到的特征点映射到高维空间中并使用经典的图像迭代配准算法LK进行配准。实验结果表明,所提算法与迭代最近点算法(ICP)、全局优化的ICP算法(Go-ICP)和PointNetLK相比,在无噪、有噪的情况下配准效果均最好。其中,在无噪情况下,与PointNetLK相比,所提算法的旋转均方误差降低了74.61%,平移均方误差降低了47.50%;在有噪声的情况下,与PointNetLK相比,所提算法的旋转均方误差降低了73.13%,平移均方误差降低了44.18%,说明所提算法与PointNetLK相比鲁棒性更强。将所提算法应用于放疗前后癌症患者人体点云模型的配准,从而辅助医生治疗,并实现了精确放疗。

关键词: 点云配准, 特征提取, 残差注意力机制, 深度学习, 放疗

Abstract:

Aiming at the problems of low accuracy and poor robustness of traditional point cloud registration algorithms and the inability of accurate radiotherapy for cancer patients before and after radiotherapy, an Attention Dynamic Graph Convolutional Neural Network Lucas-Kanade (ADGCNNLK) was proposed. Firstly, residual attention mechanism was added to Dynamic Graph Convolutional Neural Network (DGCNN) to effectively utilize spatial information of point cloud and reduce information loss. Then, the DGCNN added with residual attention mechanism was used to extract point cloud features, this process was not only able to capture the local geometric features of the point cloud while maintaining the invariance of the point cloud replacement, but also able to semantically aggregate the information, thereby improving the registration efficiency. Finally, the extracted feature points were mapped to a high-dimensional space, and the classic image iterative registration algorithm LK (Lucas-Kanade) was used for registration of the nodes. Experimental results show that compared with Iterative Closest Point (ICP), Globally optimal ICP (Go-ICP) and PointNetLK, the proposed algorithm has the best registration effect with or without noise. Among them, in the case without noise, compared with PointNetLK, the proposed algorithm has the rotation mean squared error reduced by 74.61%, and the translation mean squared error reduced by 47.50%; in the case with noise, compared with PointNetLK, the proposed algorithm has the rotation mean squared error reduced by 73.13%, and the translational mean squared error reduced by 44.18%, indicating that the proposed algorithm is more robust than PointNetLK. And the proposed algorithm is applied to the registration of human point cloud models of cancer patients before and after radiotherapy, assisting doctors in treatment, and realizing precise radiotherapy.

Key words: point cloud registration, feature extraction, residual attention mechanism, deep learning, radiotherapy

中图分类号: